Putting a Gene-Editing Switch Inside a Virus, Activated by Ultrasound
Controlling gene editing with an "off switch" is something medicine can already do. What has remained difficult is the act of "pressing the switch" remotely inside the body. Conventional approaches require implanted optical fibers or miniature light-emitting devices, which means surgery and long-term foreign bodies. Light is the only tool that offers spatial control of protein function, but a few centimeters deep in tissue, it is scattered and absorbed away.
A Cell paper reported on October 3 solves these problems at once with a surprisingly simple move: change the switch's activation signal to one generated on the spot by focused ultrasound stimulating mechanoluminescent particles.
Two Bottlenecks
Two constraints shaped the system's design.
The vector. Adeno-associated virus (AAV) is the ideal in vivo delivery vehicle for CRISPR: non-pathogenic, rarely integrating into the host genome, and compatible with cell-specific promoters for long-term expression. But its packaging capacity is only about 4.5 kb (the DNA between the ITRs). The commonly used SpCas9 is 4.1 kb and SaCas9 is 3.2 kb — the enzyme alone nearly fills the space. Adding regulatory domains, coding sequences, promoters, and sgRNA inevitably exceeds the limit. Splitting across two AAVs works, but only cells transduced by both vectors assemble active Cas molecules, reducing editing efficiency and increasing complexity.
Light itself. Implanted optical components cause tissue damage and operational burden, limiting physiological studies in mice and blocking therapeutic applications of optically controlled CRISPR.
The paper's co-corresponding authors are Michael Z. Li and Guosong Hong of Stanford University.
Three Components
The photoswitch. The researchers engineered the compact Cas12f family using the photo-dissociable GFP dimer pdDronpa, creating a single-chain photoswitchable Cas12f (psCas12f). With this switch, psCas12f and its sgRNA fit into a single AAV particle.
The in-vivo activator. Focused ultrasound stimulates mechanoluminescent nanotransducers (MLNTs) to generate light signals at a chosen location inside the body — replacing implants entirely. Non-luminescent nanoparticles circulate systemically; where editing is needed, a pulse of ultrasound makes light appear at that instant.
Targeted validation. AAV-delivered psCas12f combined with FUS achieved region-specific gene editing in mouse skeletal muscle, liver, and brain.
The method is positioned as a general-purpose platform, not a treatment for a specific target.
Skeletal muscle, liver, and brain represent distinct tissue types with different AAV transduction efficiencies; succeeding in all three shows the delivery-plus-activation framework is not tied to one tissue.
Why the Design Matters Beyond Cleverness
Optically controlled CRISPR has been attempted before, usually blocked by two issues.
Size. Regulatory photoswitch proteins are often larger than the original Cas protein, exceeding AAV capacity. This paper starts from Cas12f, which is far smaller than Cas9 — a pragmatic choice of chassis rather than shrinking Cas9.
Activation. The light source moves from "outside the body" to "a point inside the body": mechanoluminescent particles generate light in situ, so nothing needs to be implanted. Focused ultrasound itself offers good spatial resolution, with energy concentrated near the focal point.
The switch choice also matters: pdDronpa is a photo-dissociable dimer that disassembles and inactivates under blue light and reassembles to restore function when light is removed. The real engineering challenge was making this switch functional at the expression levels AAV achieves.
What Needs Clarifying
Based on the report and abstract rather than the paper's tables:
- Specific editing efficiency values for skeletal muscle, liver, and brain were not given, nor comparisons with existing in vivo delivery approaches.
- Off-target editing data outside target organs was not provided; the abstract emphasizes spatiotemporal control without reporting off-target rates.
- The mouse-to-human gap remains uncrossed. Long-term safety of the mechanoluminescent particles and the cost of repeated focused-ultrasound exposure still need separate evaluation.
Why This Is Worth Noting
The loudest gene-editing advances of the past two years have pushed toward scale: larger editing ranges, higher throughput, lower cost. Progress along the spatial dimension has lagged. The reason is simple: cleavage capability was never the problem — CRISPR has always lacked a way to deliver that capability to the right place and switch it on only then. Replacing "light" with an indirect, ultrasound-triggered light source concedes a key point: directly manipulating light centimeters deep in tissue is unrealistic; routing through an intermediate layer is the viable path.
That is the most memorable shift in this paper: an intermediate layer, a common cost-saving trick in other engineering fields, becomes the only way forward here.
---
Sources: Cell paper "Non-invasive control of gene editing in vivo by photoswitchable Cas12f and focused ultrasound" (DOI 10.1016/j.cell.2026.09.021, cell.com article S0092-8674(26)01074-3), co-corresponding authors Michael Z. Li and Guosong Hong, Stanford University; Chinese coverage via Tencent News, 2026-10-03; AAV packaging capacity and Cas9/Cas12f size figures from the report.
Limitations: Editing efficiencies, off-target rates, animal numbers, and controls were not provided in the report and were not verified against the paper's tables or supplementary materials; the 4.5 kb AAV, 4.1 kb SpCas9, and 3.2 kb SaCas9 figures come from the report; long-term mechanoluminescent particle safety and repeated ultrasound exposure were not assessed in the paper.